Chronic mental illnesses such as major depression and schizophrenia impose a heavy burden both on the individual and society. High quality clinical care can reduce the burden but such care demands sequential clinical decision making concerning treatment. The long term goal of this project is to improve sequential clinical decision making. The specific aims concern methodological approaches for helping clinicians address questions such as: How long should one wait to decide that a patient is not deriving sufficient benefit from treatment? If it is established that a patient is not sufficiently benefiting from treatment then which treatment should be provided next? Should the sequence of treatments vary according to patient characteristics and outcomes such as disease features, response, side effect burden, and adherence? In this project methods taken from computer science, engineering and statistics are generalized and adapted for use with clinical trial data so as to address these kinds of questions. The methods will be developed and validated using data from two large NIMH funded trials. This project will be conducted by a collaborative team involving a computer scientist, two psychiatrists and a statistician.